Steve: If you're looking for a knowledge representation take a look at neural propositions: http://independent.academia.edu/PiagetModeler
They employ forward, reverse, and cascade activation flow. ~PM Date: Wed, 25 Feb 2015 00:57:05 -0800 Subject: Re: [agi] Multiverse alternative to disambiguation From: [email protected] To: [email protected] PM, Ben, et al, Lakoff seems to be making the usual assortment of errors common to the field of cognitive psychology, especially: 1. Thinking because he thinks he sees SOME way of doing something, that it most work that way, 2. Thinking that our puny mental models of our own operation have anything at all to do with what really goes on behind our eyeballs, 3. Talking in generalities about hyper-complex systems without doing the neural network simulations to confirm that they might even work. It is quite difficult to make systems of bidirectional components operate in a stable manner, and 4. Not addressing the math involved, which is essential to make the quantitative computations to make something like Lakoff is talking about actually work. Continuing... On Tue, Feb 24, 2015 at 7:59 PM, Piaget Modeler via AGI <[email protected]> wrote: What was relevant from my perspective was Lakoff's neural research, I didn't see or hear any research, just speculations. rather than his metaphor research, and the fact that he used the term cascade to indicate how activationflows in many directions simultaneously. It didn't look like there was enough there to code to, but perhaps Ben has another take on this. BEN: Have you considered using something like Lakoff's methods for your work? If so, how would you represent the various entities to operate multi-directionally as Lakoff describes, and do so in a conventional uni-directional computer? Steve Date: Tue, 24 Feb 2015 18:30:19 -0800 Subject: Re: [agi] Multiverse alternative to disambiguation From: [email protected] To: [email protected] PM, Related but slightly off-point. Lakoff is concentrating on metaphors, which can be very obtuse, e.g. the common metaphor that "we have the best government that money can buy" which is a reference to our system of legalized bribery, sometimes called a bribocracy. Japanese does this MUCH more extensively, as they often include reference to characters in well-known folk stories to convey personality characteristics, like we might refer to someone moving up to a house of sticks in obtuse reference to the story of the Three Little Pigs. I presume the Japanese schmucks have simply not yet discovered Yiddish. Sure, some metaphors make sense as they are written, without having to understand the context they were first made in, but those are the easy ones. There appears from my own unpublished research to be ~20K common idioms and metaphors, which can be coded along with plain text that means the same thing without additional outside understanding. Once this has been done, a computer can simply substitute the plain text wherever it encounters an idiom or metaphor on the list. Similes, however can be "a horse of a different color". B-:D> However, buried in this may be some of what I am looking for. I'll have to make another pass to "read between the lines". This HAS been a fun post to write. Thoughts? Steve ============= On Tue, Feb 24, 2015 at 8:57 AM, Piaget Modeler via AGI <[email protected]> wrote: Lakoff's synopsis: http://www.ualberta.ca/~iclc2013/PRESENTATIONS/2013-07-12-17-54-49-george_lakoff.pdf ~PM George Lakoff refers to that as a cascade: http://georgelakoff.com/tag/cascades/ ~PM Date: Mon, 23 Feb 2015 22:46:16 -0800 Subject: [agi] Multiverse alternative to disambiguation From: [email protected] To: [email protected] Hi all, It dawned on me that disambiguation might be a really bad idea. Instead, when dealing with the uncertain meaning of a passage, suppose the passage were simply accepted for ALL of its possible meanings. Every pronoun could be ANY of the available nouns, etc. Of course most of the "possible" meanings would be complete nonsense, but let's see where this goes... There appears to be two obvious mechanisms where this would work itself out: 1. The computer would be looking for things it could relate to - things that address points of the computer's concern. The nonsensical interpretations wouldn't do this, and so would be ignored. This would work well for something like DrEliza. 2. The "flow of logic" from one sentence to the next would work for the valid interpretations, but not the invalid interpretations. Some invalid interpretations might work together, but simply letting the longest chain win would probably outperform any known method of disambiguation. Of course it is possible that when the analysis is done, the posting or letter could mean 2 or more different things. Here, it seems necessary to accept ALL defensible interpretations. Thoughts? Steve AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription -- Full employment can be had with the stoke of a pen. Simply institute a six hour workday. That will easily create enough new jobs to bring back full employment. AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription -- Full employment can be had with the stoke of a pen. Simply institute a six hour workday. That will easily create enough new jobs to bring back full employment. AGI | Archives | Modify Your Subscription ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
